NEXTPOST
You vote. I write.
Demand-driven publishing.
Most blogs guess what their readers want and find out afterwards, in the traffic numbers. This one asks first. nextpost.blog is a standing, public queue of articles I have already done the work for — you vote on which one gets written next.
Public counts. LinkedIn sign-in only to cast a vote.
Recently delivered
Voted for, then written. Newest first.
- 4 Sept 2026 The microphone that took a day — and still doesn't work cloudmigration.blog
- 30 Aug 2026 How I make AI videos with my own GPUs: 3090 vs 4090, first results privateai.blog
- 30 Aug 2026 The weights you run were probably not made by the lab that trained them privateai.blog
- 23 Aug 2026 GPU passthrough cost me nothing — until the GPUs had to talk to each other privateai.blog
- 12 Aug 2026 It didn't start today. It started 30 years ago. privateai.blog
- 12 Aug 2026 Goodbye WordPress, Hello Astro cloudmigration.blog
How it works
Everything here is already researched
The queue is not a wishlist. Every item is a piece I have already built, measured or broken something over — the writing is what is missing, not the work.
You vote for what you want to read
Sign in with LinkedIn and vote. One click, and you can take it back. The point of the sign-in is that a vote comes from a real person with a real job, which is what makes the ranking mean anything.
The votes reorder my week
I write the top of the queue next and publish it on the blog it belongs to. The vote is not a suggestion box that nobody reads — it is the actual running order.
Two lanes
One queue, feeding two blogs. Each item says where it will be published.
Private AI
Running real models on your own GPUs: serving, benchmarks, energy per token, the hardware underneath, and the parts that did not work.
Read the blog →Cloud Exit & Hardware
Leaving the hyperscalers on purpose: hypervisors, passthrough, networking, storage and what the on-prem alternative actually costs.
Read the blog →Why I built this
I have a long list of things I have measured, built or broken in my own lab, and a much shorter list of things I have found time to write up properly. Choosing between them used to be a guess. Now the people who would actually read the article get to make that call.
There is a second, more honest reason: a vote tells me what a real engineer with a real problem wants to know more about. That is far more useful than a page-view count, and it is why the sign-in is there at all — not to collect an audience, but to make the ranking mean something.
Reader polls are not new — newsletters and creator platforms have run them for years. What is unusual here is that the queue is permanent, public, spans both blogs, and every item on it is already researched. Nothing gets voted onto the list that I cannot actually deliver.